Longitudinal analysis of new information types in clinical notes

Area of Science:

• Clinical Informatics
• Natural Language Processing
• Biomedical Data Science

Background:

• Redundant information in electronic health records (EHRs) hinders secondary data use for research and patient care.
• Developing automated methods to distinguish relevant new information from redundant text is crucial.

Purpose of the Study:

• To investigate automated methods for identifying redundant versus relevant new information in clinical notes.
• To assess the utility of these methods for improving clinical information extraction systems.

Main Methods:

• Utilized automated techniques to analyze clinical reports for redundant and new information.
• Employed Unified Medical Language System (UMLS) semantic types to extract problems, medications, and laboratory data.
• Correlated automatically identified new information with manual reference standard annotations.

Main Results:

• Automated methods demonstrated high correlation with manual annotations for identifying new clinical information.
• The system successfully extracted problems, medications, and laboratory information.

Conclusions:

• Automated identification of new information in clinical notes is feasible and accurate.
• These methods can enhance clinical information extraction systems, aiding researchers and clinicians in navigating EHR data efficiently.
• Improved data extraction supports better clinical research and patient care by highlighting health status changes.